House Agents
House Agents is a specialized software suite of sub-agents designed to optimize workflows for Claude Code by preventing context window pollution. It addresses the issue of verbose data such as file search results, documentation pages, build logs, and conversation history overwhelming the main AI context. The solution employs a modular architecture where four dedicated sub-agents operate in isolated windows: House Research for file and documentation search, House Git for diff and commit analysis, House Bash for command execution and output parsing, and House MCP for tool configuration. These agents process heavy inputs and return only condensed, relevant summaries to the primary conversation, significantly reducing token usage. Production data indicates token savings ranging from 95 to 99 percent across various tasks. The system utilizes a hybrid model strategy where the main conversation uses high-performance models like Sonnet or Opus for complex decision-making, while the sub-agents run on the faster, cost-